A true AI, on par or above our thought level, is certainly an interesting concept but a very dangerous one if done badly. If someone forgets up and doesn't add some kind of inhibitor and it just goes crazy as soon as you turn it on then that'll be a problem. It gets worse if you place ranks of robots (police, military robots etc.) under its control and it eventually becomes more and more intelligent until it defects from human civilization and starts a war. Even if we can get Asimov's Laws-esque inhibitors on an AI, the intelligence issue might come up.
If it works out, however, then it can be an amazing boost to human society. Police forces are dissolved and replaced with troops of robotic police officers controlled by the AI, armies of robot soldiers fighting our wars and keeping humans safe, potential tracking of dangerous subjects throughout the entire area under the AI's watch, the list goes on and on.
